Professional Documents
Culture Documents
Manuscript received August 2, 2017; revised October 14, 2017 and K.P. Vadrevu is with the Earth Science Office, NASA Marshall Space Flight
November 27, 2017; accepted December 13, 2017. This work was supported by Center, Huntsville, AL, 35805 USA.
University of Maryland and American Society for Engineering Education V.T. Tran is with the Faculty of Information Technology, Hanoi Pedagogical
SMART scholarship. University 2, Hanoi, Vietnam.
K. Lasko and C. Justice are with the Department of Geographical Sciences,
University of Maryland, College Park, MD 20740, USA.
have also included use of texture-based classification such as thousand ha in year 2014, with the vast majority of growth
Gray Level Co-occurrence Matrix (GLCM) with promising occurring in Hanoi province [1].
results [22], [44], [45]. In addition to mapping paddy rice areal
extent, other studies have employed SAR for agricultural 1) Satellite data
phenology estimation with multi-temporal imagery [46], [47], The Sentinel-1 satellite from the European Space Agency
including studies with rice paddy fields and polarization provides C-band SAR imagery (5.4GHz) near-globally with a
variation, [48]-[55]. While some of these studies applied time- 12 day revisit time or 6 day revisit time depending on
series filters to derive accurate phenology metrics, these are not availability of Sentinel-1B imagery. The Sentinel-1 imagery is
always requisite due to the unique dynamic range of paddy rice provided as dual-polarized Interferometric Wide swath (IW)
data with vertical-transmit, vertical receive (VV), and vertical
signal from SAR [41]. The SAR signal seen in a time-series can
transmit, horizontal receive (VH) polarizations. Each
accurately capture different stages of crop growth useful for
polarization is at a nominal spatial resolution of 5x20m prior
monitoring biophysical variables as well as length of season,
to pre-processing using an open-access operational baseline
planting and harvest dates. While local farmers and observation strategy. A full time-series stack of 22 Sentinel-
stakeholders have knowledge of the crop phenology, SAR- 1A images was acquired during the 2016 growing season
based information is useful for broader national or regional (February-October) (Table 1). No Sentinel-1B imagery was
operational monitoring of crops (i.e. crop conditions, status, or available over the study area. Level-1 ground-range detected,
health) which is a major goal of international initiatives such as descending mode, IW imagery acquired from the Alaska
Asia-Rice Crop Estimation and monitoring (Asia-RiCE) project Satellite Facility (a direct mirror of ESA’s Sci-hub) were
or the Global Agricultural Monitoring (GEOGLAM) initiative processed using the free and open source Sentinel-1 toolbox.
[56], [57]. The SAR observations can survey the entire region, The ground-range detected images were processed following
whereas optical data can often be obstructed by cloud. Thus, guidelines including applying restituted orbit files, multi-look
SAR data can be useful for crop mapping and monitoring at azimuthal compressions to 20m, terrain correction using
regular intervals. SRTM 30m version 4 DEM, radiometric calibration
adjustments to correct for viewing geometry effects, and
II. STUDY AREA AND DATASETS refined lee speckle filter to reduce constructive and destructive
interference, all resulting in sigma-nought backscatter data
A. Study Area logarithmically scaled in dB [64]. Given the relatively small
The mosaic landscape of the Hanoi Capital Region includes study region with generally flat terrain, incident angle artifacts
a variety of land cover types dominated by rice agriculture, as and layover/shadow effects were relatively minimal. We
well as other small-holder croplands, urban areas, small display incident angle and SAR backscatter ranges across the
plantations, and aquaculture. The typical rice field size in the study area in Table 1. We also highlight average SAR signal
area is about 800m2 with fields routinely planted in large and the temporal dynamic across the different land cover types
collectives as necessary to facilitate irrigation. The field size is found in the study area (Fig. 1).
suitable for moderate-to-fine resolution mapping [58]. The
Hanoi Capital Region is situated within the Red River Delta, 2) Training and Validation Data
Vietnam’s oldest rice producing region, accounting for about We conducted fieldwork throughout the study area during
15% of the country’s total rice production [1]. Of all the May/June and September/October of 2016 as part of a related
different crops, rice is most prevalent and it is the dominant project on rice residue burning. The data included surveys on
crop type in the region. It has two distinct seasons: Winter- crop calendar, crop rotation, field conditions, and biomass
Spring, and Spring-Summer. Rice has three distinct stages: data. Over 900 geolocated photos were taken of paddy fields
sowing/transplanting, growth, and harvest/post-harvest, all of and non-rice areas such as aquaculture, wetlands, and other
which can be identified using satellite data [8]. In the study land cover types found within the region. These field photos
area, rice is sown or transplanted after the Tet holiday in were used for training or validation along with fine-resolution
February or early-March. Subsequently, a significant green-up imagery from Google Earth and the original Sentinel-1
is observed as the rice matures in its vegetative stage, imagery. No training data were included in the validation. We
especially after heading in April [59], [60]. During late-May to show a multi-temporal composite of the SAR data with the
June the rice is harvested, and rapidly prepared for the next training polygons overlaid (Fig. 2).
season starting in late-June or July. After the harvest in each
season (June and October) the rice residues including straw III. METHODOLOGY
and stubble, are regularly burned accounting for as much as We classified the 6 different datasets (Table 2) using the
13% of PM2.5 emissions for Vietnam; thus burning may random forest algorithm to obtain rice areal estimates and as a
constitute a significant air quality issue [58], [61]. basis to compare the datasets. These resulting datasets were
The rice fields are supported by a vast network of irrigation the basis for comparison of the different polarizations and
and drainage canals and small access roads seen throughout spatial resolutions for mapping small-holder paddy rice.
the region. While most fields practice double cropping, some Accordingly, we selected nominal 10m resolution as it’s the
fields cultivate a single crop of rice. After the harvest, single native resolution of Sentinel-1 and 20m resolution data for
rice fields remain flooded to support the growing aquaculture comparison. We addressed the following questions in our
industry [62], [63]. The study area has seen rapid expansion of study: 1) Which dataset yields the highest overall and class
aquaculture with 16.5 thousand ha in year 2000, and 46
specific accuracies and mapped areas, and are there notable assessment of the area mapped for each class. We also
differences between them? 2) What is the typical rice computed the confidence intervals for adjusted class and
phenology including start/end of season and length of growing overall accuracies to be used for comparison amongst the
season? 3) How do the resulting mapped rice area estimates different maps.
vary at the province and commune level?4) Based on
C. Spatial Analysis
landscape analysis, what are the spatial characteristics of
paddy fields and approximately how many are in the study We compared the paddy rice maps through a number of
area? metrics including Average Absolute Deviation (AAD) (1) and
Coefficient of Variation (CV) (2) to explore absolute and
A. Double and single crop rice mapping relative variation in mapped area for each of the 6 different
We employed an ensemble, machine-learning random forest datasets as follows:
algorithm for mapping single and double crop paddy rice, as ∑|𝑥− 𝑥̅ |
𝐴𝐴𝐷 = (1)
well as for evaluating the relative importance of specific input 𝑁
𝜎
data for mapping [65]. The machine-learning random forest 𝐶𝑉 = ( ) 100 (2)
𝑥̅
algorithm implemented in Scikit-learn python package, uses Where x is the rice area in hectares for one of the six rice
bootstrap aggregated sampling to build individual decision maps, 𝑥̅ is the mean rice area of all 6 datasets, and N is the
trees for classification. Each decision tree was built with a number of datasets for (1). While for (2) σ is the standard
bootstrap sample from the training data, with the unsampled deviation of all rice area for all six datasets, and 𝑥̅ is the mean
data used for out-of-bag sampling. Within the structure of a rice area for all datasets. The final value is a percentage.
tree, a random sample of the square root of the number of We compared the AAD and CV at the commune-level
predictors was chosen for each split as best candidates derived (third-level administrative subdivision) and study-area level to
from the entire predictor set. Random forest is robust against assess overall and spatial variability. In addition we computed
outliers and over-fitting, nonparametric, has high classification pixel-level thematic change for each dataset. We also
accuracy, and can yield a measure of variable importance. employed Fragstats (Version 4.2) for landscape-scale analysis
We implemented the random forest algorithm using all 22 on the most accurate resulting dataset including number of
(VV or VH) or 44 input bands (VV & VH) (Table 1) for each patches (collections of connected/adjacent paddy rice pixels)
of the six datasets separately. We populated the random forest and patch metrics useful for evaluating the size of large
with 1000 trees, where out-of-bag errors reach asymptotic connected aggregations/collections of paddy rice fields.
values [65]. The same classification technique was applied to Further, we employed mean-shift image segmentation on the
each dataset. Out of bag samples, which are randomly double-rice pixels of the time-series imagery stack to estimate
withheld from classification training were used as an indicator rice collective size (groups of fields with similar crop
of feature importance. From the 1000 trees populated in our phenology) and number of fields in the study area by
random forest, the mean decrease in classification accuracy for including typical paddy field size of 790m2 [58]. We show the
the input bands was reported. This was useful as a measure of general flowchart in Fig. 3. Mean-shift image segmentation is
feature importance in mapping paddy rice and each input band a non-parametric iterative algorithm fitting a neighborhood
can be linked to a general crop growth stage (i.e. sowing, window around each pixel, calculating the data mean in the
vegetative growth, and harvest). Training polygons were window, and shifting the neighborhood window to the mean
digitized over the single rice, double rice, and non-rice areas [68], [69]. The algorithm is useful for clustering pixels with
within each province using field photos, Sentinel-1 data, similar signal and has been used in a variety of land cover
survey information, and fine-resolution Google Earth imagery. remote sensing applications [70], [71].
B. Accuracy assessment D. Rice phenology
We performed an accuracy assessment on the resulting rice While some of the prior studies utilized spatiotemporal
maps. Random points across the study area were generated filtering to derive crop phenology metrics, we derived the
using a stratified random sampling scheme [66]. Based on the metrics based on the unique rice dynamic range combined
proportion of each resulting mapped class, a stratified random with general phenology timeframes. Sentinel-1 VH-polarized
sample of 402 total polygons for double crop rice (125), single backscatter imagery were used to estimate sowing
crop rice (41) and non-rice (236) classes were generated for /transplanting and harvest dates for both seasons of rice. Based
the 10m data. For the 20m data the same polygons were used on the unique phenology of paddy rice measured by SAR and
for comparison consistency, however, there are fewer total selecting a general timeframe based on a-priori knowledge, we
pixels due to reduced spatial resolution. We specified a 600m found local planting date coincides with the local minimum
minimum distance between polygons to prevent field overlap. value (indicative of flooding, constrained to February or
Following good practices in accuracy assessment we March), and harvest date when the local maximum backscatter
adjusted the classification accuracies as well as the mapped value is reached (indicative of peak maturity of rice just prior
rice area estimates based on the weighting from the proportion to harvest, constrained to May or June) as noted in previous
of land area for each class [67]. This weighting results in studies [8], [41], [48], [72]. The estimated range is based on
unbiased areal estimates. Based on the same, we derived the overpass dates of the Sentinel-1 satellite. The approximate
uncertainty estimates for the accuracy-adjusted areas using a length of growing season is derived by differencing the
95% confidence interval for each of our resulting mapped median date within the planting range and the harvest date
classes. The accuracy-adjusted area provides a more robust range.
Nam (5.71%), and Hung Yen (4.17%). We attribute the these are approximations dependent on the Sentinel-1 overpass
highest variation in Vinh Phuc to small-holder agriculture and dates (+-11 days). The estimated average sowing date for the
mosaicked landscape making classification difficult. The z- 2nd season is Day of Year (DoY) 175 (23rd of June) and DoY
score results for double-rice suggested more insight into 55 (24th of February) for the first season. For the harvest we
province-level variation. For VV20m Ha Nam and Hung Yen found an average DoY of 148 (27th of May) for season 1.
had the highest positive z-scores whereas all other provinces Whereas for season 2 we found estimated average harvest of
had negative Z-scores, and that VV had the most variation in DoY 279 (5th of October). These estimates are generally in
z-score across the provinces. This suggests the VV may be line with our local knowledge and field experience. A more
more sensitive to different environmental parameters such as precise estimate of length of season, harvest, and sowing could
water and moist soil underlying much of these areas. We note be obtained by including both Sentinel-1A and Sentinel-1B,
similar Z-scores for VVVH20m with the least variation across cutting the overpass time in half.
provinces in single and double crop rice. The same also goes
H. Rice landscape metrics
for VH20m with all z-scores for each dataset as positive.
Generally the 20m datasets tend to have more similar z-scores Fragstats analysis using the most accurate VVVH10m map
than the 10m datasets. produced a total of 31,934 large continuous tracts/patches of
double-rice pixels with an average area of 6.72ha (σ=59.4ha)
F. Commune-level rice variation and large standard deviation. Mean-shift image segmentation
For the entire study area the total mapped double rice area combined with fragstats analysis on the time-series stack
for each dataset is relatively similar ranging from: 208,276ha masked for double-rice pixels, suggests 796,720 rice
(VV10m) to 220,536ha (VVVH20m). However, results at the collectives (small patches of rice with similar SAR signal)
commune level (third-level administrative subdivision) with an average area of 2668m2 (σ=7250m2). The average
suggest more variation. We note that while VV10m had the field size is reported to be 790m2 (σ =625m2) [58], thus an
overall lowest double-rice mapped area, the trend is not average of 3.4 individual rice fields make up a typical rice
spatially universal as VV10m had 27 communes with more collective which follow a similar rice phenology pattern
rice than VH10m. For example, VH10m in Loung Tai had including harvest and flooding dates, due to similarity in
249ha less rice than VV10m with a percent difference of image segmentation. Based on the aforementioned average
4.60%. However, the biggest percent difference between the field size and total double-rice area from the VVVH10m
two was in Ba Vi where VV10m had 20% less rice than scenario we estimate 2.69 million double-rice paddy fields
VH10m. within the HCR.
To better evaluate spatial variation at the commune level,
I. Input band importance
we measured the AAD and CV. Across all communes for each
dataset we found rice area did not vary uniformly across The feature importance of each input band for the paddy
space, based on the AAD and CV results (Fig. 8). We found rice classification was assessed in the random forest classifier
an average CV of 4.75% for each commune. The spatial through mean decrease in accuracy (Fig. 10). We found that
pattern suggested that much of the variation proportionately the top 3 most important bands for VH(10m and 20m), and
occurred in the northern half of the study area (Vinh Phuc, VV(20m) were generally the same with bands 11, 12, and 13
northern Hanoi, and northern Bac Ninh provinces), whereas (Fig. 10). These three bands were from June and early July
communes in Hung Yen and Ha Nam had variation between (Table 1) suggesting that imagery from spring harvest and
the datasets almost always less than 2.5%. We also note that summer planting season are most important as removing one
the communes surrounding Hanoi City had relatively higher of these bands results in a mean decrease in accuracy of 3-6%
CV, but low AAD due to small amount of rice in these areas. depending on the dataset. In addition, these bands are
We attribute the spatial clustering of variation to land cover important for discriminating between single and double crop
and land use differences, and the difficulty of mapping in a rice. Other bands obtained during planting, growth, and
small-holder mosaicked landscape. Accordingly, the northern harvest, however, are also important as removing any one of
areas dominated by more wetlands, hills, and especially them still reduces overall accuracy significantly. For VV (10m
smaller rice fields intermixed with small-holder non-rice crops only) we found bands 7, 6, and 8 to be most important. These
are likely to have greater class confusion. bands are from the main vegetative growth stage for winter
rice (April). A relatively similar trend was observed in VVVH
G. Rice phenology datasets. The average local incidence angle and standard
We estimated the length of growing season, start of season, deviation over rice areas suggests minimal variation in
and end of season for rice based on the SAR time-series and incidence angle (µ = 38.63degrees, σ = 3.34degrees). Thus, it
the unique dynamic range in SAR signal found for rice is not critical to account for the incidence angle variation
paddies (Fig. 9). The results were aggregated to commune effects on backscatter in the study area (Table 1) due to
level for comparison. Across all communes we found an minimal variation. In addition, the time-series signal of double
average length of the first rice season to be 93 days, while the rice versus single rice shows it is possible to differentiate
second season was approximately 104 days. We attribute the between the two with only the first season of data. Double rice
relatively lower range in season 1 due to most fields remaining areas have relatively higher backscatter throughout the first
flooded for several months from the previous season harvest rice season (Fig. 1).
and missing SAR data for ~June 28th. In reality, the first
season length is likely slightly underestimated. In addition
aquaculture applications and land use planning. It may also be National University Hanoi) for field assistance, and William
beneficial to integrate moderate resolution Sentinel-1 SAR Salas and Nathan Torbick (Applied Geosolutions, LLC) for
imagery with high resolution UAV-obtained imagery for SAR expertise. K. Lasko thanks Michael Campbell
relating to monitoring and improving crop growth parameter (Geospatial Research Lab) for support and also thanks his PhD
estimation [80]. committee members including Christopher Justice, Krishna
Vadrevu, Ivan Csiszar, Louis Giglio, and Matthew Hansen for
VI. CONCLUSION support. Lastly, we thank the European Space Agency for the
Using a random forest classifier, we compared single and open data policy and Alaska Satellite Facility for hosting a
double crop paddy rice mapping using 6 different datasets mirror of the Sentinel-1 data.
based on varied polarizations (VV, VH, and both) as well as
different spatial resolutions (10m and 20m). Local incidence VIII. REFERENCES
angles for the complete SAR time-series dataset suggest that [1] General Statistics Office of Vietnam, 2016. Agriculture, Forestry, and
imagery can be obtained for the whole year with negligible Trade. Accessed from:
http://www.gso.gov.vn/Default_en.aspx?tabid=491.
variation reducing the need to account for angular influence on
[2] Pham, V.C., Pham, T.T.H., Tong, T.H.A., Nguyen, T.T.H. and Pham,
backscatter. The time-series Sentinel-1 SAR imagery was able N.H., 2015. The conversion of agricultural land in the peri-urban areas
to estimate start of season, end of season, and length of season of Hanoi (Vietnam): patterns in space and time. J. Land Use Sci., 10(2),
for each rice crop at the commune level with promising pp.224-242.
results. Rice mapping estimates for the 6 datasets suggested [3] Whitcraft, A.K., Vermote, E.F., Becker-Reshef, I. and Justice, C.O.,
the most accurate dataset to be VVVH10m (93.5% OA), 2015. Cloud cover throughout the agricultural growing season: Impacts
on passive optical earth observations. Rem. Sens. Enviro., 156, 438-447.
whereas 20m datasets generally had lower overall, user’s and
[4] Paudyal, D.R. and Aschbacher, J., 1993. Land-cover separability studies
producer’s accuracies compared to the 10m datasets. Thematic of filtered ERS-1 SAR images in the tropics. In Geoscience and Remote
change between the datasets suggested the least spatial Sensing Symposium, IGARSS'93. Better Understanding of Earth
variation in classes to be between VH and VVVH (10m and Environment., International 1216-1218.
20m). The most common difference was between double crop [5] Aschbacher, J., Pongsrihadulchai, A., Karnchanasutham, S., Rodprom,
rice and non-rice areas. 20m datasets overestimated rice land C., Paudyal, D.R. and Le Toan, T., 1995. Assessment of ERS-1 SAR
area more compared with 10m datasets, as pixels were too data for rice crop mapping and monitoring. In Geoscience and Remote
Sensing Symposium, 1995. IGARSS'95.'Quantitative Remote Sensing for
large to capture small roads and ditches within paddy fields. Science and Applications', International Vol. 3, 2183-2185.
The 20m datasets, however, could be well-representative of [6] Takeuchi, S. and Suwanwerakamtorn, R., 1996. Monitoring of Land
not only rice land cover, but especially rice land use. Cover Conditions in Paddy Fields Using Multitemporal SAR Data. Int.
Province-level variation in mapped rice areas suggested the Archives of Photogramm. And Rem. Sens., 31, 683-688.
most occurring with VV20m, and also more variation in single [7] Chakraborty, M., Panigrahy, E.S. and Sharma, S.A., 1997.
rice than double rice based on the coefficient of variation. Discrimination of rice crop grown under different cultural practices
using temporal ERS-1 synthetic aperture radar data. ISPRS J.
Moderately-high spatial variation was found at the commune- Photogramm. Remote Sens., 52(4), 183-191.
level with the maximum variation in rice area as 20% with [8] Le Toan, T., Ribbes, F., Wang, L.F., Floury, N., Ding, K.H., Kong, J.A.,
VV10m and VH10m. This spatial variation could be critical Fujita, M. and Kurosu, T., 1997. Rice crop mapping and monitoring
for spatially-explicit land cover mapping applications. using ERS-1 data based on experiment and modeling results. IEEE
Average absolute deviation and coefficient of variation results Transactions on Geoscience and Remote Sensing, 35(1), 41-56.
suggest communes in the northern portion of the study area [9] Liew, S.C., Kam, S.P., Tuong, T.P., Chen, P., Minh, V.Q. and Lim, H.,
1998. Application of multitemporal ERS-2 synthetic aperture radar in
and those immediately surrounding Hanoi City have the most
delineating rice cropping systems in the Mekong River Delta,
proportional variation. This is important as these areas are Vietnam. IEEE Trans. Geosci. Rem. Sens., 36(5), 1412-1420.
subject to peri-urban expansion and important for air quality [10] Yun, S., Cuizhen, W., Xiangtao, F. and Hao, L., 1997, August.
studies as areas closest to the city could have the highest Estimation of rice growth status using RADARSAT data. In Geoscience
impact on urban air quality. and Remote Sensing, 1997. IGARSS'97. Remote Sensing-A Scientific
Overall our study compared six different datasets (VV10m, Vision for Sustainable Development., 1997 IEEE International (Vol. 3,
1430-1432). IEEE.
VV20m, VH10m, VH20m, VVVH10m, and VVVH20m) for
[11] Ribbes, F., 1999. Rice field mapping and monitoring with RADARSAT
mapping paddy rice, evaluated importance of input bands from data. Int. J. Rem. Sens., 20(4), 745-765.
different dates for classification, estimated rice phenology and [12] Panigrahy, S., Manjunath, K.R., Chakraborty, M., Kundu, N. and
calculated rice landscape metrics. The results from our study Parihar, J.S., 1999. Evaluation of RADARSAT standard beam data for
provide useful insights for crop mapping/monitoring efforts identification of potato and rice crops in India. ISPRS J. Photogramm.
using SAR datasets in general and particularly within the Red Remote Sens., 54(4), 254-262.
River Delta of Vietnam. [13] Shao, Y., Fan, X., Liu, H., Xiao, J., Ross, S., Brisco, B., Brown, R. and
Staples, G., 2001. Rice monitoring and production estimation using
multitemporal RADARSAT. Rem. Sens. Enviro., 76(3), 310-325.
VII. ACKNOWLEDGEMENT [14] Lee, K. S., and S. I. Lee. "Assessment of post-flooding conditions of rice
This work was supported by the University of Maryland, fields with multi-temporal satellite SAR data." Int. j. Rem. sens. 24, no.
17 (2003): 3457-3465.
Council on the Environment Green Fund Fellowship. K.
[15] Suga, Y., Takeuchi, S., Oguro, Y. and Konishi, T., 2000. Monitoring of
Lasko was partially supported by the Science, Mathematics rice-planted areas using space-borne SAR data. Int. Arch. of
and Research for Transformation (SMART) scholarship by the PHOTOGRAM. and Rem. Sens., 33(B7/4; PART 7), pp.1480-1483.
American Society for Engineering Education. The authors [16] Inoue, Y., Kurosu, T., Maeno, H., Uratsuka, S., Kozu, T., Dabrowska-
thank Thanh Nhat Thi Nguyen and Hung Quang Bui (Vietnam Zielinska, K. and Qi, J., 2002. Season-long daily measurements of
multifrequency (Ka, Ku, X, C, and L) and full-polarization backscatter [35] Son, N.T., Chen, C.F., Chen, C.R. and Minh, V.Q., 2017. Assessment of
signatures over paddy rice field and their relationship with biological Sentinel-1A data for rice crop classification using random forests and
variables. Rem. Sens. of Enviro., 81(2), 194-204. support vector machines. Geocarto Int., pp.1-15.
[17] Chen, C. and McNairn, H., 2006. A neural network integrated approach [36] Steele-Dunne, S.C., McNairn, H., Monsivais-Huertero, A., Judge, J.,
for rice crop monitoring. Int. J. Rem. Sens., 27(7), pp.1367-1393. Liu, P.W. and Papathanassiou, K., 2017. Radar remote sensing of
[18] Salas, W., Boles, S., Li, C., Yeluripati, J.B., Xiao, X., Frolking, S. and agricultural canopies: A review. IEEE J-STARS, 10(5), pp.2249-2273.
Green, P., 2007. Mapping and modelling of greenhouse gas emissions [37] Whitcomb, J., Moghaddam, M., McDonald, K., Kellndorfer, J. and
from rice paddies with satellite radar observations and the DNDC Podest, E., 2009. Mapping vegetated wetlands of Alaska using L-band
biogeochemical model. Aquatic conservation: Marine and freshwater radar satellite imagery. Canadian J. of Rem. Sens., 35(1), 54-72.
ecosystems, 17(3), 319-329. [38] White, L., Brisco, B., Dabboor, M., Schmitt, A. and Pratt, A., 2015. A
[19] Zhang, Y., Wang, C., Wu, J., Qi, J. and Salas, W.A., 2009. Mapping collection of SAR methodologies for monitoring wetlands. Rem.
paddy rice with multitemporal ALOS/PALSAR imagery in southeast sens., 7(6), pp.7615-7645.
China. Int. J. Rem. sens., 30(23), 6301-6315. [39] Wang, W., Yang, X., Li, X., Chen, K., Liu, G., Li, Z. and Gade, M.,
[20] Lam-Dao, N., Apan, A., Young, F.R., Le-Van, T., Le-Toan, T. and 2017. A Fully Polarimetric SAR Imagery Classification Scheme for
Bouvet, A., 2007, November. Rice monitoring using ENVISAT-ASAR Mud and Sand Flats in Intertidal Zones. IEEE Trans. Geosci. Rem.
data: preliminary results of a case study in the Mekong River Delta, Sens., 55(3), pp.1734-1742.
Vietnam. In Proceedings of the 28th Asian Conference on Remote [40] Shiraishi, T., Motohka, T., Thapa, R.B., Watanabe, M. and Shimada, M.,
Sensing (ACRS 2007). Geoinformatics Center (Klong Luang). 2014. Comparative assessment of supervised classifiers for land use–
[21] Bouvet, A., Le Toan, T. and Lam-Dao, N., 2009. Monitoring of the rice land cover classification in a tropical region using time-series palsar
cropping system in the Mekong Delta using ENVISAT/ASAR dual mosaic data. IEEE J-STARS, 7(4), pp.1186-1199.
polarization data. IEEE Trans. on Geosci. Rem. Sens., 47(2), 517-526. [41] Torbick, N., Salas, W., Chowdhury, D., Ingraham, P. and Trinh, M.,
[22] Jia, K., Li, Q., Tian, Y., Wu, B., Zhang, F. and Meng, J., 2012. Crop 2017. Mapping rice greenhouse gas emissions in the Red River Delta,
classification using multi-configuration SAR data in the North China Vietnam. Carbon Management, 8(1), pp.99-108.
Plain. Int.l J. Rem. Sens., 33(1), pp.170-183. [42] Waske, B. and Braun, M., 2009. Classifier ensembles for land cover
[23] Nguyen, D.B., Clauss, K., Cao, S., Naeimi, V., Kuenzer, C. and Wagner, mapping using multitemporal SAR imagery. ISPRS Journal of
W., 2015. Mapping rice seasonality in the Mekong Delta with multi-year Photogrammetry and Remote Sensing, 64(5), pp.450-457.
Envisat ASAR WSM data. Rem. Sens., 7(12), pp.15868-15893. [43] Sonobe, R., Tani, H., Wang, X., Kobayashi, N. and Shimamura, H.,
[24] Pei, Z., Zhang, S., Guo, L., McNairn, H., Shang, J. and Jiao, X., 2011. 2014. Random forest classification of crop type using multi-temporal
Rice identification and change detection using TerraSAR-X TerraSAR-X dual-polarimetric data. Rem. Sens. Lett., 5(2), 157-164.
data. Canadian J. Rem. Sens., 37(1), 151-156. [44] Zhang, H., Li, Q., Liu, J., Du, X., Dong, T., McNairn, H., Champagne,
[25] Gebhardt, S., Huth, J., Nguyen, L.D., Roth, A. and Kuenzer, C., 2012. A C., Shang, J. and Liu, M., 2017. Object-based crop classification using
comparison of TerraSAR-X Quadpol backscattering with RapidEye multi-temporal SPOT-5 imagery and textural features with a Random
multispectral vegetation indices over rice fields in the Mekong Delta, Forest classifier. Geocarto International, pp.1-40.
Vietnam. Int. j. rem. sens., 33(24), pp.7644-7661. [45] Uhlmann, S. and Kiranyaz, S., 2014. Integrating color features in
[26] Inoue, Y., Sakaiya, E. and Wang, C., 2014. Capability of C-band polarimetric SAR image classification. IEEE Transactions on
backscattering coefficients from high-resolution satellite SAR sensors to Geoscience and Remote Sensing, 52(4), pp.2197-2216.
assess biophysical variables in paddy rice. Rem. Sens. Enviro., 140, [46] Lopez-Sanchez, J.M., Cloude, S.R. and Ballester-Berman, J.D., 2012.
pp.257-266. Rice phenology monitoring by means of SAR polarimetry at X-
[27] Nelson, A., Setiyono, T., Rala, A.B., Quicho, E.D., Raviz, J.V., band. IEEE Trans. Geosci. Rem. Sens., 50(7), 2695-2709.
Abonete, P.J., Maunahan, A.A., Garcia, C.A., Bhatti, H.Z.M., Villano, [47] Jiao, X., Kovacs, J.M., Shang, J., McNairn, H., Walters, D., Ma, B. and
L.S. and Thongbai, P., 2014. Towards an operational SAR-based rice Geng, X., 2014. Object-oriented crop mapping and monitoring using
monitoring system in Asia: Examples from 13 demonstration sites across multi-temporal polarimetric RADARSAT-2 data. ISPRS J. Photogram.
Asia in the RIICE project. Rem. Sens., 6(11), pp.10773-10812. Rem. Sens., 96, pp.38-46.
[28] Jiao, X., Kovacs, J.M., Shang, J., McNairn, H., Walters, D., Ma, B. and [48] Koppe, W., Gnyp, M.L., Hütt, C., Yao, Y., Miao, Y., Chen, X. and
Geng, X., 2014. Object-oriented crop mapping and monitoring using Bareth, G., 2013. Rice monitoring with multi-temporal and dual-
multi-temporal polarimetric RADARSAT-2 data. ISPRS J. Photogram. polarimetric TerraSAR-X data. Int. J. Appl. Earth Obs. Geoinfo., 21,
Rem. Sens., 96, pp.38-46. pp.568-576.
[29] Hoang, H.K., Bernier, M., Duchesne, S. and Tran, Y.M., 2016. Rice [49] Erten, E., Rossi, C., Yüzügüllü, O. and Hajnsek, I., 2014, July.
Mapping Using RADARSAT-2 Dual-and Quad-Pol Data in a Complex Phenological growth stages of paddy rice according to the BBCH scale
Land-Use Watershed: Cau River Basin (Vietnam). IEEE J-STARS. and SAR images. In Geosci. and Rem. Sens. Symp. (IGARSS), 2014
g, 9(7), 3082-3096. IEEE International (pp. 1017-1020). IEEE.
[30] Asilo, S., de Bie, K., Skidmore, A., Nelson, A., Barbieri, M. and [50] Erten, E., Rossi, C. and Yüzügüllü, O., 2015. Polarization impact in
Maunahan, A., 2014. Complementarity of two rice mapping approaches: TanDEM-X data over vertical-oriented vegetation: The paddy-rice case
Characterizing strata mapped by hypertemporal MODIS and rice paddy study. IEEE Geosci. Rem. Sens. Lett., 12(7), 1501-1505.
identification using multitemporal SAR. Rem. Sens., 6(12), 12789- [51] Lopez-Sanchez, J.M., Vicente-Guijalba, F., Ballester-Berman, J.D. and
12814. Cloude, S.R., 2014. Polarimetric response of rice fields at C-band:
[31] Karila, K., Nevalainen, O., Krooks, A., Karjalainen, M. and Analysis and phenology retrieval. IEEE Trans. Geosci. Rem.
Kaasalainen, S., 2014. Monitoring changes in rice cultivated area from Sens., 52(5), pp.2977-2993.
SAR and optical satellite images in Ben Tre and Tra Vinh Provinces in [52] Küçük, Ç., Taşkın, G. and Erten, E., 2016. Paddy-rice phenology
Mekong Delta, Vietnam. Rem. Sens., 6(5), pp.4090-4108. classification based on machine-learning methods using multitemporal
[32] Torbick, N., Chowdhury, D., Salas, W. and Qi, J., 2017. Monitoring co-polar X-band SAR images. IEEE J-STARS, 9(6), pp.2509-2519.
Rice Agriculture across Myanmar Using Time Series Sentinel-1 [53] de Bernardis, C.G., Vicente-Guijalba, F., Martinez-Marin, T. and Lopez-
Assisted by Landsat-8 and PALSAR-2. Rem. Sens., 9(2), p.119. Sanchez, J.M., 2015. Estimation of key dates and stages in rice crops
[33] Nguyen, D.B., Gruber, A. and Wagner, W., 2016. Mapping rice extent using dual-polarization SAR time series and a particle filtering
and cropping scheme in the Mekong Delta using Sentinel-1A data. Rem. approach. IEEE J-STARS, 8(3), pp.1008-1018.
Sens. Lett., 7(12), pp.1209-1218. [54] Xie, L., Zhang, H., Wu, F., Wang, C. and Zhang, B., 2015. Capability of
[34] Mansaray, L.R., Zhang, D., Zhou, Z. and Huang, J., 2017. Evaluating rice mapping using hybrid polarimetric SAR data. IEEEJ-STARS, 8(8),
the potential of temporal Sentinel-1A data for paddy rice discrimination pp.3812-3822.
at local scales. Rem. Sens. Lett., 8(10), pp.967-976.
[55] Yuzugullu, O., Erten, E. and Hajnsek, I., 2015. Rice growth monitoring [76] Boschetti, M., Nelson, A., Nutini, F., Manfron, G., Busetto, L., Barbieri,
by means of X-band co-polar SAR: Feature clustering and BBCH M., Laborte, A., Raviz, J., Holecz, F., Mabalay, M.R.O. and Bacong,
scale. IEEE Geosci. Rem. Sens. Lett., 12(6), 1218-1222. A.P., 2015. Rapid assessment of crop status: An application of MODIS
[56] Oyoshi, K., Tomiyama, N., Okumura, T., Sobue, S. and Sato, J., 2016. and SAR data to rice areas in Leyte, Philippines affected by Typhoon
Mapping rice-planted areas using time-series synthetic aperture radar Haiyan. Remote Sensing, 7(6), pp.6535-6557.
data for the Asia-RiCE activity. Paddy and Water Environment, 14(4), [77] Belgiu, M. and Drăguţ, L., 2016. Random forest in remote sensing: A
pp.463-472. review of applications and future directions. ISPRS J. Photogramm.
[57] Whitcraft, A.K., Becker-Reshef, I. and Justice, C.O., 2015. A framework Remote Sens., 114, pp.24-31.
for defining spatially explicit earth observation requirements for a global [78] Loosvelt, L., Peters, J., Skriver, H., De Baets, B. and Verhoest, N.E.,
agricultural monitoring initiative (GEOGLAM). Remote Sensing, 7(2), 2012. Impact of reducing polarimetric SAR input on the uncertainty of
pp.1461-1481. crop classifications based on the random forests algorithm. IEEE
[58] Lasko, K., Vadrevu, K., Tran, V., Ellicott, E., Nguyen, T., Bui, H. and Transactions on Geoscience and Remote Sensing, 50(10), pp.4185-4200.
Justice, C., 2017. Satellites may underestimate rice residue and [79] Rodriguez-Galiano, V.F., Chica-Olmo, M., Abarca-Hernandez, F.,
associated burning emissions in Vietnam. Enviro. Res. Lett., 12(8), Atkinson, P.M. and Jeganathan, C., 2012. Random Forest classification
085006. of Mediterranean land cover using multi-seasonal imagery and multi-
[59] Xiao, X., Boles, S., Liu, J., Zhuang, D., Frolking, S., Li, C., Salas, W. seasonal texture. Rem. Sens. Environ., 121, pp.93-107.
and Moore, B., 2005. Mapping paddy rice agriculture in southern China [80] Uto, K., Seki, H., Saito, G. and Kosugi, Y., 2013. Characterization of
using multi-temporal MODIS images. Rem. Sens. Enviro., 95(4), rice paddies by a UAV-mounted miniature hyperspectral sensor
pp.480-492. system. IEEE J-STARS, 6(2), pp.851-860.
[60] Kontgis, C., Schneider, A. and Ozdogan, M., 2015. Mapping rice paddy
extent and intensification in the Vietnamese Mekong River Delta with
dense time stacks of Landsat data. Rem. Sens. Enviro., 169, 255-269. Kristofer Lasko received the B.S. degree in
[61] Duong, P.T. and Yoshiro, H., 2015. Current Situation and Possibilities geographical sciences and M.S. degree in
of Rice Straw Management in Vietnam. Accessed from: geospatial information science from the
http://www.jsrsai.jp/Annual_Meeting/PROG_52/ResumeC/C02-4.pdf. University of Maryland, College Park, MD,
[62] Dey M.M. and Ahmed, M., 2005. Aquaculture—food and livelihoods USA in 2012 and 2014 respectively. He is
for the poor in Asia: a brief overview of the issues. Aquaculture
Economics & Management, 9(1-2), pp.1-10.
currently pursuing the PhD degree focusing
[63] Ottinger, M., Clauss, K. and Kuenzer, C., 2017. Large-Scale Assessment on estimating rice residue burning and
of Coastal Aquaculture Ponds with Sentinel-1 Time Series Data. Rem. emissions variation through integrating field
Sens., 9(5), p.440. and remote sensing data with a focus in the Vietnam region.
[64] Zuhlke, M., Fomferra, N., Brockmann, C., Peters, M., Veci, L., Malik, J. He has worked as a research intern and project lead at two
and Regner, P., 2015, December. SNAP (Sentinel Application Platform) different NASA locations. He subsequently worked for the
and the ESA Sentinel 3 Toolbox. In Sentinel-3 for Science
University of Maryland as a Geospatial Scientist while
Workshop (Vol. 734, p. 21).
[65] Breiman, L., 2001. Random forests. Machine learning, 45(1), pp.5-32. pursuing his master’s degree. He is currently a graduate
[66] Congalton, R.G. and Green, K., 2008. Assessing the accuracy of research assistant at the University of Maryland and on a
remotely sensed data: principles and practices. CRC press. scholarship from the American Society for Engineering
[67] Olofsson, P., Foody, G.M., Herold, M., Stehman, S.V., Woodcock, C.E. Education.
and Wulder, M.A., 2014. Good practices for estimating area and
assessing accuracy of land change. Rem. Sens. Enviro., 148, 42-57
Krishna P. Vadrevu received the PhD
[68] Fukunaga, K. and Hostetler, L., 1975. The estimation of the gradient of a
density function, with applications in pattern recognition. IEEE
degree in ecology and remote sensing in
Transactions on information theory, 21(1), pp.32-40. 2000 while working at the Indian Space
[69] Comaniciu, D. and Meer, P., 2002. Mean shift: A robust approach Research Organization (ISRO), National
toward feature space analysis. IEEE Transactions on pattern analysis Remote Sensing Center, Hyderabad, India.
and machine intelligence, 24(5), pp.603-619. His thesis focused on biodiversity and forest
[70] Su, T. and Zhang, S., 2017. Local and global evaluation for remote characterization using remote sensing
sensing image segmentation. ISPRS Journal of Photogrammetry and
Remote Sensing, 130, pp.256-276.
techniques.
[71] Senthilnath, J., Omkar, S.N., Mani, V., Tejovanth, N., Diwakar, P.G. From 2001 to 2003, he worked as a Post-doctoral
and Shenoy, A.B., 2012. Hierarchical clustering algorithm for land cover Researcher and 2004-2009 as a Research Scientist at the Ohio
mapping using satellite images. IEEE J-STARS, 5(3), pp.762-768. State University, USA. From 2010-2016 he worked as an
[72] Choudhury, I. and Chakraborty, M., 2006. SAR signature investigation Associated Research Professor with the Department of
of rice crop using RADARSAT data. Int. J. Rem. Sens., 27(3), pp.519- Geographical Sciences, University of Maryland college park,
534.
USA. He joined the NASA Marshall Space Flight Center in
[73] Chen, C.F., Son, N.T., Chen, C.R., Chang, L.Y. and Chiang, S.H., 2016.
Rice crop mapping using Sentinel-1A phonological metrics. ISPRS- Huntsville, Alabama, USA as a remote sensing scientist
International Archives of the Photogrammetry, during July, 2016. He also serves as the lead scientist for
[74] Nguyen, D., Wagner, W., Naeimi, V. and Cao, S., 2015. Rice-planted NASA South/Southeast Research Initiative
area extraction by time series analysis of ENVISAT ASAR WS data (www.sari.umd.edu) and is on the editorial board for five
using a phenology-based classification approach: A case study for Red different peer-reviewed journals.
River Delta, Vietnam. The International Archives of Photogrammetry,
Remote Sensing and Spatial Information Sciences, 40(7), p.77.
[75] Hoang, K.H., Bernier, M., Duchesne, S. and Tran, M.Y., 2014, July.
Classification of rice fields in a complex land-use watershed in Northern
Vietnam using RADARSAT-2 data. In Geoscience and Remote Sensing
Symposium (IGARSS), 2014 IEEE International (pp. 1501-1503). IEEE.
Avg Angle Std Angle Avg angle Std Angle Avg angle Std Angle
Band Date (Double Rice) (Double Rice) (Single Rice) (Single Rice) (Not Rice) (Not Rice)
1 02/03/2016 38.630 3.343 39.304 3.328 39.547 6.414
2 02/15/2016 38.639 3.347 39.314 3.326 39.547 6.414
3 02/27/2016 38.640 3.349 39.316 3.327 39.543 6.414
4 03/10/2016 38.629 3.341 39.306 3.328 39.550 6.412
5 03/22/2016 38.640 3.350 39.326 3.322 39.539 6.411
6 04/03/2016 38.644 3.351 39.334 3.326 39.538 6.409
7 04/15/2016 38.640 3.343 39.325 3.322 39.545 6.411
8 04/27/2016 38.644 3.345 39.327 3.322 39.546 6.411
9 05/09/2016 38.637 3.344 39.322 3.325 39.543 6.410
10 05/21/2016 38.644 3.353 39.331 3.337 39.542 6.408
11 06/02/2016 38.624 3.337 39.306 3.327 39.547 6.412
12 06/14/2016 38.634 3.342 39.315 3.325 39.546 6.412
13 07/08/2016 38.632 3.347 39.303 3.327 39.547 6.410
14 07/20/2016 38.632 3.340 39.313 3.326 39.548 6.412
15 08/01/2016 38.620 3.336 39.302 3.327 39.547 6.412
16 08/13/2016 38.627 3.339 39.303 3.327 39.550 6.411
17 08/25/2016 38.627 3.339 39.304 3.327 39.550 6.412
18 09/06/2016 38.626 3.336 39.307 3.327 39.553 6.412
19 09/18/2016 38.629 3.340 39.307 3.327 39.548 6.412
20 09/30/2016 38.629 3.340 39.310 3.325 39.543 6.412
21 10/12/2016 38.638 3.346 39.325 3.327 39.542 6.410
22 10/24/2016 38.640 3.344 39.325 3.322 39.543 6.411
Local incidence angle, θ, variation in SAR imagery per mapped land cover class for VV&VH 10m
dataset: Minimal variation through time in all categories. ‘Not rice’ areas have higher standard
deviation due to mountainous areas.
REFERENCE REFERENCE
VV Class Double rice Single rice Non-rice Total Adj. User's Acc. (C.I) Area (ha) Adj. User's
VV Class Double rice Single rice Non-rice Total Acc. Mapped Area (ha)
Double rice 389 3 14 406 95.8% (0.10%) 208,276
Double rice 116 2 8 126 92.1% (0.42%) 214,903
Single rice 0 145 1 146 99.3% (0.11%) 4,908
Single rice 0 48 0 48 100% (0.00%) 5,450
Non-rice 99 13 919 1031 89.1% (0.06%) 594,549
Non-rice 29 2 298 329 90.6% (0.17%) 587,376
Total 488 161 934 1583 807,733 Total 145 52 306 503 807,729
Adj. Producer's 77.8% 98.7% 97.5%
Acc (0.17%) 35.0% (0.58%) (0.02%) Adj. Producer's Acc. 79.3% (0.55%) 43.8% (1.91%) (0.10%)
confusion matrices with associated pixels per class, mapped class area, and adjusted accuracies with confidence intervals based on calculated unbiased areal
estimates for a) 10m data; and b) 20m data.
Time-series SAR signal over different land cover types and polarizations in the study area.
Figure 3 includes each SAR rice map and a pixel count of double-rice agreement between the 6 datasets on the right side.
Copyright 2017 IEEE. Author submitted final version.
5
Area (ha)
Z-score
Z-score
Double rice and single cropped rice province-level z-scores indicating deviation from the mean.
Average absolute deviation and coefficient of variation for double rice area per commune with province boundaries overlaid. Results
suggest most AAD in communes with high rice mapped areas, while CV results suggest communes in the north tended to have
higher percentage variation likely attributed to mapped area difficulties in more mosaicked landscapes with smaller fields.
Copyright 2017 IEEE. Author submitted final version.
Figure 9
Commune-level rice phenology
Commune-level sowing and harvest day of year (DoY) ranges based on Sentinel-1 signal. Reason for
Copyright 2017 IEEE. Author submitted final version.
minimal variation in summer sowing is due to missing data for ~06/28/2016, thus actual range is bigger.
Figure 10
Mean Decrease in Accuracy
Mean decrease in out of bag accuracy (%): suggests bands from all parts of rice stage
(planting/growth/harvest) are important, with bands from second season planting and first season harvest
Copyright
generally most critical. Note for VVVH 2017 IEEE.
graphs, Author submitted
bands 1-22 are final version.
VV and bands 23-44 are VH.